Big Tech keeps $3 trillion of AI exposure off balance sheets through accounting footnotes
The biggest technology companies in the world have collectively stacked up roughly $3 trillion in AI-related financial commitments that don’t show up as liabilities on their balance sheets. That’s five times what these same companies spent on capital expenditures over the past year.
Nine major tech firms, including Alphabet, Amazon, Meta, Microsoft, and Oracle, have quietly amassed this mountain of obligations through unstarted data-center leases and purchase agreements for chips, equipment, and infrastructure. The commitments are technically disclosed, just buried in financial footnotes where most investors rarely look.
The footnote iceberg
Of the roughly $3 trillion total, approximately $1.2 trillion consists of data-center leases that haven’t started yet. The remaining $1.9 trillion is locked into purchase commitments for the chips and equipment needed to power AI workloads.
To put that in perspective, the trailing 12-month capital expenditures for these nine companies came in at around $600 billion. So the off-balance-sheet pile is about five times larger than what’s currently flowing through their income statements.
This is all perfectly legal under GAAP, the accounting framework that governs US financial reporting. Certain types of leases and purchase agreements can be disclosed in footnotes rather than recorded as liabilities until they’re activated.
Some of these companies have also used special-purpose vehicles, essentially standalone legal entities designed to keep debt off the parent company’s consolidated balance sheet. Meta’s Hyperion project, financed by Blue Owl Capital, is one example of this kind of structure in action.
A separate analysis from Nikkei Asia estimated that just five hyperscalers had accumulated $1.65 trillion in similar hidden obligations, a figure that exceeded their $1.35 trillion in total reported debt.
The numbers are moving fast
Alphabet’s contractual obligations hit $811 billion as of June 30, 2026. Three months earlier, that number was $332 billion. A $479 billion increase in a single quarter is the kind of jump that would normally set off alarm bells, except that it’s tucked into a footnote rather than splashed across the headline financials.
Oracle’s off-balance-sheet commitments have exploded to roughly $273 billion, representing a 30-fold increase over four years.
Meta’s share of these obligations sits at an estimated $420 billion, nearly five times its reported debt.
Why this matters for investors
The problem is one of legibility. Most investors assess a company’s leverage by looking at the balance sheet. If a significant chunk of financial obligations lives in footnotes, the standard metrics, debt-to-equity ratios, net debt calculations, leverage multiples, all understate the true picture.
Analysts have flagged that hyperscalers are experiencing increased leverage and mounting pressure on free cash flow as these commitments begin converting into actual spending. The gap between commitments made and cash generated creates refinancing risk, particularly if AI revenue growth doesn’t keep pace with the infrastructure buildout.
